Two innovative modelling approaches in order to forecast consumption of blast furnace gas by hot blast stoves

Abstract

Abstract The online optimization of the use of process off gases in integrated steelworks can greatly contribute to increase the sustainability of the steel production. A correct management of these gases could allow both the reduction of natural resources exploitation (e.g. natural gas) and of the facility's environmental impact. However, in order to achieve an almost complete use of these gases, it is fundamental to forecast their production and consumption according to the production plan and to use such forecasting to optimize the gases distribution inside the network by considering possible interactions. According to these needs, this paper presents two models, which allow forecasting the consumption of blast furnace gas by some major consumers: the hot blast stoves. Due to the almost regular operation of these plants, two kinds of models can be applied: an Echo State Network-based model, which is more complex and sensitive to the variations of the operating practices and a simpler switch model, which does not require training and is very easy to use. Both models provide good results and the user can interchangeably exploit them.